Bossanyi 2003

WIND ENERGY
Wind Energ. 2003; 6:229–244 (DOI: 10.1002/we.95)
Review Wind Turbine Control for

Article Load Reduction
E. A. Bossanyi,∗ Garrad Hassan and Partners Ltd., St Vincent’s Works, Silverthorne Lane, Bristol
BS2 0QD, UK
Key words: This article reviews techniques for the control of wind turbines during power production.
wind turbine Pitch control is used primarily to limit power in high winds, but it also has an important
control; variable effect on structural loads. Particularly as turbines become larger, there is increasing interest
speed; pitch
control; torque in designing controllers to mitigate loads as far as possible. Torque control in variable-speed
control; turbines is used primarily to maximize energy capture below rated wind speed, and to limit
closed-loop the torque above rated, but it can also be used to reduce certain loads. The design of the
control; control algorithms is clearly of prime importance. Additional sensors such as accelerometers
damping; load and load sensors can also help the controller to achieve its objectives more effectively. By
alleviation
controlling the pitch of each blade independently, it is also possible to achieve important
further reductions in loading. It is important to be able to quantify the benefits of any new
controller. Although computer simulations are useful, field trials are also vital. The variability
of the real wind means that particular care is needed in the design of the trials. Copyright 
2003 John Wiley & Sons, Ltd.
Introduction
Blade pitch control is primarily used to limit the aerodynamic power in above-rated wind speeds, in order to
keep the turbine within its design limits. Some optimization of energy capture below rated is also possible.
In variable-speed turbines, generator torque control is used primarily to limit the transmission torque in
above-rated winds, and to control the rotor speed below rated in order to maximize energy capture in this
region.
The algorithms used for controlling pitch and torque need careful design. In addition to their effectiveness
in meeting these primary objectives, the control algorithms can also have a major influence on the loads
experienced by the wind turbine. Clearly the algorithms must be designed so as to prevent excessive loading,
but it is possible to go further by designing them with load reduction as an explicit objective. As the size of
wind turbines increases, and as cost reduction targets encourage lighter and hence more flexible and dynamic
structures, these aspects of controller design become increasingly important.
A very basic controller might consist of a classical PI or PID algorithm acting on a single measured signal
(generator speed or power output) to generate a pitch demand. For variable-speed turbines a torque demand
is generated independently from a speed–torque look-up table. This basic scheme can be greatly improved
in a number of ways. This article covers the following possibilities:
ž joint control of pitch and torque to improve the trade-off between energy and loads;
ž using torque control to damp out torsional resonances, e.g. in the drive train;
ž adding a nacelle-mounted accelerometer to help the controller to reduce fore–aft tower vibration;
ž individual pitch control to reduce asymmetrical loadings.
Ł
Correspondence to: E. A. Bossanyi, Garrad Hassan and Partners Ltd., St Vincent’s Works, Silverthorne Lane, Bristol
BS2 0QD, UK. E-mail: bossanyi@garradhassan.com
Copyright  2003 John Wiley & Sons, Ltd.

230 E. A. Bossanyi
The first two of these are now routinely used in the industry. With the increasing use of tall, flexible towers
and the availability of cheap and reliable accelerometers, the third option is of increasing interest. Individual
pitch control is also likely to be of major interest as turbines become larger and more flexible, and also as
suitable sensors are now becoming more available.
Classical Designs
PI (proportional and integral) and PID (proportional, integral and derivative) controllers are widely used
throughout industry and are a good starting point for many wind turbine control applications. A PID controller
can be written in terms of the Laplace variable s (similar to a differentiation operator) as

Ki Kd s
yD C Kp C x 1
s 1 C s
where x is the input error signal to be corrected and y is the control action. Ki , Kp and Kd are the integral,
proportional and derivative gains. The time constant prevents the derivative term from becoming large at
high frequency, where it could respond excessively to signal noise. Kd is zero in a PI controller.
Fixed-speed Pitch Controllers

For a fixed-speed pitch-regulated turbine, x is the difference between the measured power and the demanded
or rated power, and y is the demanded pitch angle. The integral term in (1) ensures that the mean value of
x will be zero, otherwise y would increase or decrease indefinitely. Below rated the output y is limited to
the fine pitch position as x becomes negative. While y is on the limit, the integral term must be prevented
from growing indefinitely or ‘winding up’, otherwise it would take a long time before the pitch demand starts
increasing after x becomes positive again.
In order to maximize energy production below rated, the fine pitch limit may itself be varied slowly as a
function of average power to optimize aerodynamic efficiency as the tip speed ratio varies.
Variable-speed Controllers
For a variable-speed pitch-regulated turbine above rated a similar scheme is often used to regulate the rotor
speed to the desired value, while the generator torque or power is held constant. Now x is the difference
between the measured rotational speed (usually at the generator) and the demanded or rated rotational speed,
and y once again is the pitch angle.
Below rated the pitch is forced to the fine pitch limit as before, but now the generator torque can be varied
in order to control the generator speed. Maximum aerodynamic efficiency is achieved at a certain tip speed
ratio which maximizes the power coefficient Cp . Since the rotor speed is then proportional to the wind
speed V, the power increases with V3 and the torque with V2 . A simple calculation shows that the generator
torque demand Qd should be set proportional to the square of the measured generator speed ωg using the
relationship
R5 Cp 2
Qd D ωg 2
23 G3
in order to achieve peak Cp , where R is the rotor radius, is the air density and G is the gearbox ratio.
This follows from the definition of the power coefficient (power D Qd ωg D 12 R2 Cp V3 and tip speed
ratio ( D ωg /G)R/V). This relationship is shown schematically in Figure 1 as line BC. Equation (2) can
be used quite successfully to control generator torque below rated. Although it will not track optimum
Cp perfectly for reasons explained below, it is usually entirely satisfactory. Actually, the torque demand
from equation (2) should be reduced by the amount of any mechanical losses, so that the correct quadratic
relationship is maintained between rotor aerodynamic torque and rotor speed. The losses may vary with speed,
so ideally it is the overall system efficiency which should be optimized, not just the aerodynamic efficiency.
Copyright  2003 John Wiley & Sons, Ltd. Wind Energ. 2003; 6:229–244
Wind Turbine Control 231
Equation (2) depends on the air density and is therefore sometimes recalculated depending on the site altitude.
However, the energy capture is not too sensitive to this torque–speed relationship; in fact, it is possible in
practice to keep the torque demand slightly above optimum, which will reduce the rotor speed while losing
very little energy. This can be worthwhile to reduce aeroacoustic noise, since this noise is highly dependent on
the rotor speed, and is of particular importance in moderate wind speeds when this torque–speed relationship
is in use.
In practice, acoustic noise or other design constraints mean that the maximum allowable rotor speed is
reached at a relatively low wind speed. As the wind speed increases further, it is desirable to increase the
torque and power without any further speed increase, in order to capture more energy from the wind. The
simplest strategy is to implement a torque–speed ramp: line CD in Figure 1. Then, in order to prevent the
torque and pitch controllers from interfering with one another, the speed set-point for the pitch controller is
set a little higher, at point E in Figure 1. If the speed set-point were at D, then there would constantly be
power dips during operation in above-rated wind speeds, whenever the speed fell transiently below D, while
in below-rated conditions there would be nothing to prevent the pitch controller from acting unnecessarily.
Variable Speed—PI Torque Control

It would be an improvement if the torque–speed trajectory A-B-C-D-E in Figure 1 could be changed to
A-B1-C1-E. The turbine would then stay close to optimum Cp over a wider range of wind speeds, giving
slightly higher energy capture for the same maximum operating speed.1 The vertical sections A-B1 and C1-E
can be achieved by using a PI controller for the torque demand, in response to the generator speed error with
the set-point at B1 or C1. Transitions between constant-speed and optimum Cp operation are conveniently
handled by using the optimum Cp curve as the upper torque limit of the PI controller when operating at
B1, or the lower limit when at C1. The set-point flips between B1 and C1 when the measured speed crosses
the mid-point between B1 and C1. Despite this step change in set-point the transition is completely smooth,
because the controller will be forced onto the optimum Cp limit curve both before and after the transition.
This logic can easily be extended to implement ‘speed exclusion zones’, to avoid speeds at which blade
passing frequency would excite, for example, the tower resonance, by introducing additional speed set-points
and some logic for switching between them—see lines FG and HJ in Figure 1. When the torque demand
18 m/s
16 m/s
Increasing
14 m/s pitch
12 m/s D E
Generator torque
10 m/s
Optimum Cp
(fine pitch)
8 m/s
6 m/s
G C1
H
4 m/s
C
B1 B F J
A
Generator speed
Figure 1. Torque–speed curve for variable-speed operation (schematic)
232 E. A. Bossanyi
exceeds G for a certain time, the set-point ramps smoothly from F to H. Then, if it falls below J, the set-point
ramps back again.
Another advantage of PI control of the torque is that the ‘compliance’ of the system can be controlled.
Controlling to a steep ramp (CD in Figure 1) can be quite harsh in that the torque demand will be varying
rapidly up and down the slope. A PI controller, on the other hand, can be tuned to achieve a desired level
of ‘softness’. With high gain the speed will be tightly controlled to the set-point, requiring large torque
variations. Lower gains will result in more benign torque variations, while the speed is allowed to vary more
around the set-point.
Variable Speed—Controlling Pitch and Torque Together

In order to use point C1 as the speed set-point for both the torque and pitch controllers, it is necessary to
decouple the two. One technique is to arrange some switching logic which ensures that only one of the control
loops is active at any one time. Thus below rated the torque controller is active and the pitch demand is fixed
at fine pitch, while above rated the pitch controller is active and the torque demand is fixed at rated. This
can be done with fairly simple logic, although there will always be occasions when the controller is caught
briefly in the ‘wrong’ mode. For example, if the wind is just below rated but rising rapidly, it might be useful
to start pitching the blades a little before the torque demand reaches rated. If the pitch does not start moving
until the torque reaches rated, it then has to move some way before it starts to control the acceleration, and
a small overspeed may result.
A more satisfactory approach is to run both control loops together, but to couple them together with terms
which drive one or the other loop into saturation when far above or below the rated wind speed. Thus most
of the time only one of the controllers is active, but they can be made to interfere constructively when close
to the rated point.
A useful method is to include a torque error term in the pitch PID in addition to the speed error. Above
rated, since the torque demand saturates at rated, the torque error will be zero, but below rated it will be
negative. An integral term will bias the pitch demand towards fine pitch, preventing the pitch controller from
acting in below-rated winds, while a proportional term may help to start the pitch moving a little before the
torque reaches rated if the wind speed is rising rapidly.
It is also necessary to prevent the torque demand from dropping when operating well above rated wind
speed. Here a useful strategy is a ‘ratchet’, which prevents the torque demand from falling while the pitch is
not at fine. This can also smooth over brief lulls in the wind around rated, using the rotor kinetic energy to
avoid transient power drops.
Effect of Turbulence on Peak Cp Tracking

For tracking peak Cp below rated in a variable-speed turbine, the quadratic algorithm of equation (2) works
well and gives smooth, stable control. However, in turbulent winds the large rotor inertia prevents it from
changing speed fast enough to follow the wind, so rather than staying on the peak of the Cp curve it will
constantly fall off either side, resulting in a lower mean Cp . This problem is clearly worse for heavy rotors,
and also if the Cp – curve has a sharp peak.
It is possible to manipulate the generator torque to cause the rotor speed to change faster when required,
so staying closer to the peak of the Cp curve. One way to do this is to differentiate the measured speed and
subtract a term proportional to rotor acceleration from the torque demand. This has the effect of reducing the
apparent rotor inertia, making the torque tracking more responsive.1 A number of other ways have also been
proposed for improving the peak Cp tracking.2,3 However, Holley et al.3 showed (for a particular case) that a
perfect Cp tracker would capture 3% more energy below rated by demanding huge power swings of š three
or four times rated power, which is clearly impractical. There is therefore usually no reason to adopt any of
these more dynamic approaches.
Controller Tuning
Clearly the choice of controller gains is crucial to the performance of the controller. With too little overall
gain the turbine will wander around the set-point, while too much gain can make the system completely
unstable. Inappropriate combinations of gains can cause structural responses to become excited.
A linearized model of the turbine dynamics is useful for tuning the parameters of a PI or PID controller.
This allows various techniques to be used for rapidly evaluating the performance and stability of the control
algorithm. Detailed non-linear simulations using a three-dimensional turbulent wind input should then be used
to verify the design before it is implemented on the real turbine.
A wind turbine is basically non-linear. However, it is possible to linearize the dynamics about any chosen
operating point. For below-rated PI speed control using demanded torque, which can be quite slow and gentle,
the linearized model can be very simple. It must include at least the rotational dynamics of the drive train.
For pitch control the aerodynamics and some of the structural dynamics can be critical. The linearized model
should contain at least the dynamics:
ž rotor and generator rotation,

ž drive train torsion,
ž generator response,
ž tower fore–aft vibration,
ž power or speed transducer response,
ž pitch actuator response,
as well as a description of the aerodynamics of the rotor in terms of partial derivatives of torque and thrust with
respect to pitch angle, wind speed and rotor speed. The thrust is important as it affects the tower dynamics,
which couple strongly with the pitch control.
With this linear model it is then possible to vary the gains and rapidly calculate various ‘measures of
performance’ which help to evaluate the performance of the controller with those gain settings. Useful
measures of performance include:
ž gain and phase margins, to indicate how close the system is to instability;
ž the crossover frequency (at which the open-loop gain crosses unity), as a measure of the responsiveness of
the controller;
ž the positions of the closed-loop ‘poles’ of the system, which indicate how well various resonances will be
damped;
ž step responses, illustrating how, for example, the pitch angle and tower motion respond to a step change in
wind speed;
ž frequency responses, showing, for example, how much the pitch responds at critical frequencies such as
the blade passing frequency or the drive train resonant frequency, or how much the tower will be excited
by the wind.
With experience it is possible by examining measures such as these to converge rapidly on a controller tuning
which will work well in practice.
Gain Scheduling
Close to rated, since the fine pitch angle is selected to maximize power, it follows that the sensitivity of
aerodynamic torque to pitch angle is very small. Thus a much larger controller gain is required here than at
higher wind speeds, where a small change in pitch can have a large effect on torque. Frequently the torque
sensitivity changes almost linearly with pitch angle and so can be compensated for by varying the overall gain
of the controller linearly with the inverse of the pitch angle. Such a modification of gain with operating point
is termed a ‘gain schedule’. However, the thrust sensitivity varies in a different way, and so because of the
234 E. A. Bossanyi
importance of tower dynamics it may be necessary to modify the gain schedule to ensure good performance
in all winds. It is therefore necessary to create linear models of the system corresponding to several different
operating points between rated and cut-out wind speed, and to choose a gain schedule which ensures that the
above performance measures are satisfactory over the whole range.
Pitching to Stall
Most pitch-controlled turbines pitch to feather: as the wind increases, the pitch is increased, leading edge
into wind, to reduce the angle of attack and hence also the lift, so limiting the aerodynamic torque. It is also
possible to limit the torque by pitching the other way, increasing the angle of attack to cause the blade to stall
(‘pitching to stall’). This also increases the drag and hence the thrust loading on the turbine. However, both
the torque and thrust become more stable, varying more slowly with pitch angle. This means that although
the thrust loads are higher, they vary less, so fatigue loads may actually be reduced. Also, much smaller
pitch actions are needed to control the torque. This means that it is possible to use very simple, almost
‘open-loop’ algorithms to control the pitch; for example, pitching at constant rate whenever the power or
speed crosses certain thresholds. However, these algorithms are generally not amenable to analytical design
methods and usually need to be tuned by trial and error. The PI/PID approach described above can also be
used successfully for the case of pitching to stall. It provides an analytical design method and is likely to
lead to smoother control, with much gentler pitch action than in the case of pitch to feather. The fact that
the steady state pitch schedule is double-valued, with the pitch first decreasing with wind speed and then
increasing again, does not present a problem, since the gain still always has the same sign. It does make it
more difficult to implement a gain schedule, since it is not possible to know the operating point purely from
the pitch angle, but as the gain schedule is much less critical than in pitch to feather, this is rarely a problem.
If necessary, the nacelle anemometer may be used to help distinguish between the high-wind and low-wind
operating points corresponding to any measured pitch angle.
This approach to PID design of pitch-to-stall controllers has been shown to give excellent and very benign
control performance in detailed simulation studies. The main drawback is the uncertainty which remains in
the theoretical understanding of stalled rotor aerodynamics.
Damping of Resonances
The control algorithm can have a major effect on the loads on certain parts of the turbine structure. It is
important to take this into account when designing it. For example, pitch control is used primarily to regulate
the aerodynamic torque, but changes in pitch also have a major effect on the thrust load. This in turn affects
the blade out-of-plane bending moments. However, the thrust also drives the fore–aft motion of the tower.
This in turn affects the relative wind speed seen by the blades, which then feeds back into the pitch control
via the aerodynamic torque. This is strong feedback which has a major effect on the stability of the pitch
control system. It is easy to design a pitch controller which exacerbates the tower vibration or even makes it
completely unstable, with important consequences for the tower base loads.
Controlling Tower Vibration

The first tower fore–aft vibrational mode is essentially very lightly damped, exhibiting a strong resonant
response which can be maintained at quite a high level even by the small amount of excitation which is
naturally present in the wind. The strength of the response depends critically on the small amount of damping
which is present, mostly aerodynamic damping from the turbine rotor. The pitch control action modifies the
effective damping of that mode. In designing the pitch controller, it is therefore important to avoid reducing
that damping, and if possible to increase it. This is why the linear model described above must include the
fore–aft tower dynamics. A root locus plot will then show two lightly damped tower poles close to the
imaginary axis. As part of the tuning process, these poles must be pushed as far to the left as possible to
increase the damping. It is also useful to observe the peak of the closed-loop frequency response from wind
speed to tower velocity. The peak should be as low as possible.
In order to achieve this, it is sometimes useful to use a filter tuned to the tower resonant frequency in series
with the PID controller. The filter modifies the magnitude and phase of the pitch response at that frequency
in such a way as to increase the tower damping. A general second order filter of the form
s2 C 21 ω1 s C ω12
3
s2 C 22 ω2 s C ω22
can be tuned to achieve this effect. However, a more successful approach is to use an accelerometer to provide
information about the tower vibration to the controller, as explained below.
Drive Train Torsional Vibration

A typical drive train can be considered to consist of a rotor inertia and a high-speed shaft inertia (mainly
generator and brake disc), separated by a torsional spring. Sometimes it is important to consider also the
coupling of the torsional mode with the first rotor in-plane collective mode, in which case the drive train can
be approximated by three inertias and two torsional springs. In some cases the coupling to the second tower
side-to-side mode, which is characterized by large angular displacements at the tower top, is also important.
In a fixed speed turbine the induction generator slip curve essentially acts like a strong damper, and so the
torsional mode of the drive train is well damped. In a variable-speed turbine operating at constant generator
torque, however, there is very little damping for this mode, which can therefore lead to very large torque
oscillations at the gearbox, effectively negating one of the principal advantages of variable-speed operation.
Although it may be possible to provide some damping mechanically, for example by means of appropriately
designed rubber mounts or couplings, there is a cost associated with this. Another solution, which has been
successfully adopted on many turbines, is to modify the generator torque control to provide some damping.
Instead of demanding a constant generator torque above rated, a small ripple at the drive train frequency
is added on, with its phase adjusted to counteract the effect of the resonance and effectively increase the
damping. A highpass or bandpass filter of the form
2ωs1 C s
G 4
s2 C 2ωs C ω2
acting on the measured generator speed can be used to generate this additional ripple. The frequency ω must
be close to the resonant frequency which is to be damped. The time constant can sometimes be used to
compensate for time lags in the system, or to adjust the phase of the response. A root locus plot is again very
useful for tuning the filter parameters. However, because of coupling between the drive train torsion and the
rotor in-plane and tower side–side vibrations, it is difficult to construct a linear model analytically. However,
numerical approaches are now available4 which present an easy means of constructing suitable models.
It is important for the bandpass filter to be reasonably narrow so that it does not respond strongly at other
frequencies, such as multiples of the blade passing frequency. Of course, if the resonant frequency nearly
coincides with one of these multiples, say 6P, then the resonance will be very difficult to control because it
will be strongly excited.
Two or even three such bandpass filters may be placed in parallel if other resonances such as rotor collective
in-plane vibration or tower side–side vibration are also significantly excited. Torque control can help to damp
these resonances in just the same way as the drive train torsional resonance.
Figure 2 shows some simulation results for a variable-speed turbine operating in simulated three-
dimensional turbulence. A large drive train resonance can be seen to be building up. Although the power and
generator torque are smooth, the gearbox would be very badly affected. The effect of introducing a damping
filter as described above is also shown. It almost completely damps out the resonance without increasing the
electrical power variations. This is because the torque ripple needed to damp the resonance is actually very
small, because the amount of excitation is small.
236 E. A. Bossanyi
No damping With damping

700 700
power [kW]
power [kW]
Electrical
Electrical
600 600
500 500
400 400
300 300
torque [kNm]
torque [kNm]
Gearbox
Gearbox
200 200
100 100
0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Time [s] Time [s]
Figure 2. Effect of a drive train damping filter
Measuring Tower Acceleration

So far, the control algorithms considered have acted on a measured signal from a single sensor, to generate
pitch and torque demand outputs. In principle, however, it may be possible for the controller to perform a
better job if signals from additional sensors are made available to provide more information. In this case it
is important that any gain in performance outweighs any possible loss of reliability, since all sensors have a
certain probability of failure, and sensor faults are a significant source of turbine downtime.
The importance of tower dynamics to the pitch controller has already been stressed, and this is in fact one
of the main constraints on the design of the algorithm. However, it is relatively easy to measure tower motion
directly by using an accelerometer mounted in the nacelle. This provides an extra signal which allows the
controller to distinguish more successfully between the effect of genuine wind speed changes and of tower
motion on the measured power or generator speed. By using this extra signal, it is in fact possible to reduce
tower loads significantly without adversely affecting the quality of speed or power regulation.
The tower dynamics can be modelled approximately as a second-order system exhibiting damped simple
harmonic motion, i.e.
MRx C DPx C Kx D F C F 5
where x is the tower displacement and F is the applied force, which in this case is predominantly the rotor
thrust. F is the additional thrust caused by pitch action. We pcan equate M with the tower modal mass and
K with the modal stiffness, such that the tower frequency is K/Mrad s1 . The damping term D is small.
The effective damping can clearly be increased if F is proportional to Px . Clearly it is easier to measure
acceleration than velocity, so the tower acceleration would have to be integrated to provide a measure of xP .
A suitable gain for F can be estimated from the partial derivative from pitch to thrust, ∂F/∂ˇ, where ˇ is
the pitch angle, in order to achieve any particular additional damping Dp :
∂F Dp
υF D υˇ D Dp xP , υˇ D xP 6
∂ˇ ∂F/∂ˇ
Clearly this is a rather simplified analysis. Sometimes an additional lead–lag filter may be useful to adjust
the phase of the accelerometer feedback term, or a notch filter may be required to prevent unwanted feedback
from other components of tower acceleration, for example at blade passing frequency. Figure 3 shows the
results of a simulation with and without such an acceleration feedback term using three-dimensional turbulent
wind input. Clearly this technique is capable of increasing the tower damping substantially, almost eliminating
the resonant response and significantly reducing tower loads. The only drawback is the increase in pitch
No tower damping With tower damping

10 10
8 8
6 6
4 4
2 2
0 0
-2 -2
-4 -4
-6 -6
150 200 250 300 150 200 250 300
Time [s] Time [s]
Upper lines: tower top displacement (in) Lower lines: pitch rate (deg/s)
Figure 3. Effect of tower acceleration feedback
actuator activity required to achieve this damping. Blade loads are also reduced, and there is almost no
impact on the quality of power and speed regulation (not shown).
Individual Pitch Control

Another aspect of control which would require additional sensors is the use of individual pitch control. On
a large turbine the wind speeds seen by each blade at any instant may differ significantly, so it would be
desirable to control the pitch of each blade independently to take account of this. This possibility has been
suggested many times over the years, a recent example being by Caselitz et al.5 In order to achieve any useful
benefit, there must be some measurement available which can distinguish what is happening at the different
blades, so that the controller can generate appropriate demand signals for each.
The simplest measure which could be used is the rotor azimuth angle. There are some effects (wind shear,
tower shadow, upflow and shaft tilt) which cause a systematic azimuth-dependent variation in the aerodynamic
conditions at a point on the blade. In principle the pitch of each blade could be modified as a function of
rotor azimuth in order to reduce the loading variations caused by these effects. In practice, however, this is
not a useful approach, because the wind speed variations across the rotor at any instant are dominated by
stochastic variations due to turbulence.
However, if the asymmetrical loading on the rotor can be measured in some way, then it is possible to
achieve some very significant load reduction using individual pitch control. The wind speed variations over the
rotor disc result in a large once-per-revolution, or 1P, component in the blade loads, together with harmonics
of this frequency, i.e. 2P, 3P, 4P and so on. With a three-bladed rotor these load components will be 120°
out of phase between the three blades, with the result that the hub and the rest of the structure will experience
the harmonics at 3P, 6P, etc., but 1P and the other harmonics will tend to cancel out.
This cancellation relies on assumptions of stationarity and linearity, but as turbines become larger with
respect to the length scales of the turbulence, these assumptions become less valid. For example, if blade 1
sees a gust as it passes top dead centre, the gust will have changed before blade 2 reaches the same position.
This means that the asymmetric loads resulting from the 1P and other harmonics no longer cancel out, and
load components at these frequencies can contribute very significantly to fatigue loads on the hub, shafts, yaw
bearing, tower, etc. The 1P load components are particularly significant on large turbines, and in principle it
should be possible to reduce these by means of individual blade pitch action at the 1P frequency, 120° out
of phase at the three blades.
238 E. A. Bossanyi
In order to do this, a reliable measure of the asymmetrical loading across the rotor is required. Traditional
strain gauges are not considered sufficiently reliable for this type of application, but with the development
of more advanced sensors in recent years, such as optical strain gauges and solid state strain measurement
devices, this approach is becoming feasible. It is also possible that accelerometers may be able to provide the
necessary information.
Two further barriers to the use of individual pitch control have also recently been overcome: the complexity
of the algorithms required and the analytical means to design them. Modern simulation codes have also
advanced to the point where they can be used to test such controllers under realistic conditions of simulated
three-dimensional turbulence. The possibility to use these codes to generate appropriate linearized models
numerically4 has also greatly facilitated the design of the necessary algorithms, since linear models are
required which take into account not only the rotational and structural dynamics and aerodynamics of the
turbine in a uniform wind field, but also the effect of asymmetrical wind speed variations and individual pitch
actions on the various loads.
Since this is a multivariable control problem, in which several inputs (including measured loads) are
simultaneously processed to generate three pitch actuator demands, initial work concentrated on the use of
so-called ‘LQG’ or linear–quadratic–gaussian control design techniques, these being among the simplest of
the so-called ‘modern’ or model-based control design methods which are directly applicable to multivariable
problems. This has led to some very successful results being demonstrated in detailed simulations.2 However,
the development of complex multivariable controllers in this way is far from straightforward, and the resulting
algorithms can be of very high order, requiring a large amount of processing on each controller time step. It
is also difficult to guarantee robustness: the controller must still be able to perform satisfactorily if the real
turbine differs somewhat from the model used for the control design, or if measured signals are contaminated
with noise, etc.
Subsequent work has been very successful in refining the design techniques to the point where excellent
performance has been obtained with greatly reduced model orders. Furthermore, simulations have been used
to demonstrate that the performance is not unduly degraded by imperfections in the turbine model or by
signal noise.
The best results have been obtained by decoupling the collective from the differential or 1P pitch action.
The collective pitch action, which is the same for all three blades, is calculated from the measured rotational
speed by means of a standard classical PI-based controller. Nacelle accelerometer feedback, as described
above, may be added if required. A zero-mean differential pitch action (different for each blade) is then
superimposed on the collective pitch demand to reduce the differential loads. The differential pitch action
requires a multivariable controller with at least two inputs (measurements) and two outputs. Although there
are three blades, the three pitch demands can be considered to consist of a collective pitch demand and two
independent differential demands. A useful approach is the d–q axis representation borrowed from three-
phase electrical machine theory,6 in which three blade root load signals are transformed into a mean value and
variations about two orthogonal axes (the ‘direct’ and ‘quadrature’ axes), which could represent the vertical
and lateral directions for example. Differential pitch ‘outputs’ in the d and q axes are then calculated, and a
reverse transformation provides the differential demands for the three blades. An LQG controller of relatively
low order can generate the d–q axis pitch demands from the d–q axis loads.
More recently, however, it has been shown that it is possible to treat the d and q axes as being almost
independent. This means that conventional classical design techniques can be applied to generate a single-
input, single output controller which can be applied separately to the d axis and the q axis. A conventional
PI controller in series with a simple filter provides very satisfactory control action. In practice there is some
interaction between the two axes, but this can be accounted for by introducing a simple azimuthal phase
shift into the d–q axis transformation, i.e. adding a constant offset to the rotor azimuth angle used in the
transformation. The complete controller now consists essentially of three simple PI controllers, some filters
in series if required, and simple rotational transformations to change between fixed and rotating axes.
This approach has yielded results comparable to the LQG approach. Some loads are reduced slightly less,
while others are reduced somewhat more. The resulting pitch activity is very similar. Furthermore, it has been
shown that it is possible to use a variety of different sensors with equal effectiveness. Results are presented
in this article using load sensors at the blade roots, on the hub or low-speed shaft, or at the yaw bearing.
With the PI approach it is particularly straightforward to switch from one set of sensors to another—all that
is required is a slight change of gain—and the resulting performance is very similar in each case.
Another advantage is that the implementation of the PI-based controllers is very much more straightforward
than for the LQG type of algorithm, which means that such algorithms are much more likely to be adopted
by control engineers in practice.
Details of both the LQG and PI approaches are presented in another article.7 Some results of detailed
simulation modelling using three-dimensional simulated turbulent wind fields are presented in Figures 4–7.
These results are based on a generic 75 m diameter, 2 MW offshore variable-speed turbine developed as part
of a CEC-funded project concerned with design recommendations for offshore wind turbines (“RECOFF”,
contract number ENK5-CT-2000-00 322).
Each simulation covered a 10 min period, using the same three-component turbulent wind field in each
case to drive the simulation. The mean wind speed was 13 m s1 and the turbulence intensity was 18Ð9% in
the longitudinal direction, 14Ð8% laterally and 10Ð6% vertically. The sample time histories shown below are
excerpts from these simulations, while the spectra and fatigue loads are calculated from the full 10 min.
Figure 4 illustrates the typical magnitude of the 1P pitch action which is required during operation around
rated wind speed. Clearly this represents a considerable increase in pitch actuator duty compared with a
conventional controller, particularly as some differential pitch action continues to be useful even in below-
rated winds, where significant load reductions may still occur without any significant loss of energy. However,
apart from a possible increase in wear and the need to take account of heat dissipation in the actuators, this
is unlikely to require major changes in the design of pitch actuators.
Figure 5 shows spectra of some of the key bending moment loads: at the blade root in the out-of-plane
direction, on the shaft, and at the yaw bearing. Several differential pitch controllers are shown, namely LQG
with blade root load sensors, and PI with each of blade root, shaft or yaw bearing load sensors. The different
differential pitch controllers give very similar results; in fact, for the blade root and shaft sensors the results
are nearly indistinguishable. The results are taken from 10 min simulations with the same three-component
turbulent wind in each case, around rated wind speed. For the blade and rotating shaft loads the large 1P
peak in the conventional case is virtually eliminated by differential pitch control. The yaw bearing loads are
in a co-ordinate system which is not rotating with the blades and are therefore dominated by a low-frequency
peak, representing the asymmetry in the wind field which is the cause of the 1P loading on the rotating
components. The effect of the differential pitch control therefore is to cut out this low frequency peak.
Blade 1 Blade 2 Blade 3 Collective

pitch controller
12
10
8
Pitch angle [deg]
6
4
2
0
-2
-4
-6
-8
180 190 200 210 220 230 240
Time [s]
Figure 4. Typical pitch angle variations close to rated wind speed
240 E. A. Bossanyi
Blade root bending moment: out of plane

9.0e + 11
Spectral density [Nm2/Hz]
1.0e + 11
1.0e + 10
1.0e + 09
1.0e + 08
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Shaft bending moment (My)

1.0e + 12
1.0e + 11
1.0e + 10
1.0e + 09
1.0e + 08
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Yaw bearing yaw moment (Mz)

7.0e + 11
1.0e + 11
1.0e + 10
1.0e + 09
1.0e + 08
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Frequency [Hz]
Collective LQG PI, blade PI, shaft PI, yaw
pitch sensors sensors bearing
sensors
Figure 5. Load spectra
Figures 6(a) and 6(b) present some typical time histories of these loads (only the LQG case is shown, but
others are similar), while Figure 7 shows the damage equivalent loads, which are a measure of the equivalent
fatigue damage caused by each load taking into account the fatigue properties of the material. S–N slopes of
(a) Collective pitch controller

3000
2500
2000
1500
[kNm]
1000
500
0
-500
-1000
-1500
0 50 100 150 200 250 300
Time [s]
Differential pitch controller
3000
2500
2000
1500
[kNm]
1000
500
0
-500
-1000
-1500
0 50 100 150 200 250 300
Time [s]
Blade root out of plane moment (upper trace) and
Shaft moment (lower trace)
(b) Collective pitch controller
1500
1000
500
[kNm]
-500
-1000
-1500
0 50 100 150 200 250 300
Time [s]
Differential pitch controller
1500
1000
500
[kNm]
-500
-1000
-1500
0 50 100 150 200 250 300
Time [s]
Yaw moment (upper trace) and
nodding moment (lower trace)
Figure 6. (a) Sample time histories of rotating loads; (b) Sample time histories of yaw bearing loads
242 E. A. Bossanyi
S-N slope = 4
1400
1200
1000
800
kNm
600
400
200
Yaw
0 Yaw bea
h bea ring M
p itc ds
Sha ring z
ve ft M
cti e loa ds Bla y My
oll
e lad ft loa ds de
roo
C gb ha loa QG tO
sin gs r in
g
al, L /P
tia
lu sin be
a e nti
lu fer
ren tia aw Dif
Dif
fe ren gy
iffe sin
D lu
tia
ren
D iffe
S-N slope = 10
1800
1600
1400
1200
kNm
1000
800
600
400
200
Yaw
0 Yaw bea
Sha bea ring M
h ring z
p itc ds ft M My
cti
ve loa ds Bla y
e l oa ds de
lle lad aft oa roo
Co gb s h gl QG tO
/P
sin ing ar i
n
al,L
lu s be nti
tia lu e
ren tia ya
w fer
fe ren ing Dif
Dif iffe us
D al
re nti
D iffe
Figure 7. Reduction in fatigue loads
4 and 10 have been used, which are typical for steel and composite materials respectively. The differential
pitch control produces a dramatic reduction in fatigue loading for the blades and shaft. For the yaw bearing,
it is only the low frequency loads which are reduced, so the effect on fatigue is more modest, since only a
relatively small number of large cycles are affected.
Evaluating Control Performance

A vital aspect of the development of new control algorithms is the assessment of their effectiveness. This can
be done to some extent using computer simulation, provided the models used for the turbine dynamics and
wind turbulence are adequate. In recent years, computer simulations of wind turbine behaviour have advanced
to the point where they can provide a very effective and reliable means of assessing the performance of a
wind turbine controller. However, it is also essential to evaluate any new technique using field trials. Because
of the variability of the real wind, systematic comparison of different techniques in the field must be carried
out very carefully, and few satisfactory examples of this can be found in the literature.
Simulation Methods
Some very powerful simulation models of wind turbines are now available, which are capable of modelling the
aerodynamics and system dynamics, as well as the wind input, in sufficient detail to allow controllers to be
evaluated and compared with some confidence. As well as a detailed representation of the aerodynamic
behaviour, including stall hysteresis for pitch-to-stall turbines, such models should ideally include the
following turbine features:
ž drive train rotational and torsional dynamics;

ž generator response;
ž blade vibrational dynamics, at least up to the first in-plane collective mode;
ž tower vibrational dynamics, preferably up to the second side-to-side mode;
ž power or speed transducer response;
ž pitch actuator response.
The wind input should include:
ž wind shear, tower shadow and upflow;

ž stochastic turbulent wind variations;
ž spatial variation of turbulence, to allow rotational sampling by the blades;
ž ideally, all three components of turbulence.
Field Trials
Despite the power and reliability of some of the simulation models now available, there is no substitute for
field trials in real wind conditions. The variability of the wind makes it particularly difficult to carry out field
trials repeatably and reliably, particularly if the effectiveness of two or more alternative controllers is to be
compared.
A good way to do this is to compare the different controllers by switching repeatedly between them at
intervals of a few minutes, and recording appropriate statistics over a long period of time covering a wide
range of wind conditions. This approach was used for the comparison of LQG, PI and extended classical
controllers referred to above for a 300 kW turbine,2 and also by Knudsen et al.8 in the comparison of PI and
H1 controllers on a 400 kW turbine.
In order to assess that the full objectives of the controller are being realized, it is important to measure an
appropriate selection of loads, and not just the basic variables such as power, speed and pitch which define
the primary objective of the controller.
Conclusions
Classical methods based on PI and PID algorithms are a good starting point for many aspects of closed-loop
controller design for fixed- and variable-speed turbines. However, as turbines become larger and more flexible,
244 E. A. Bossanyi
it is increasingly important not only to consider the effect that the controller has on component loads, but
even to design the controller with load reduction as part of the primary objective.
Terms can be introduced into the controller to help to damp resonances, such as drive train torsion
in variable-speed turbines. Additional sensors can be helpful too—an accelerometer can be effective in
controlling tower resonance.
For variable speed turbines, attention to detail in the interaction of pitch and torque controllers can
significantly improve energy capture without any compromise on loads.
Individual pitch control has potential for very significant load reduction but is not yet commercially proven.
The ability to measure differential loads reliably is important here.
Advanced controller design methods can offer an explicit mathematical formulation for the design of
controllers with multiple objectives, including load reduction. Such controllers have been used on commercial
turbines to a limited extent. However, this article illustrates the finding that, with careful design, much simpler
PI-based controllers can be developed which can achieve the same level of performance and are much more
likely to be adopted in practice.
Finally, the importance of careful evaluation of controller designs is stressed. The use of simulations and
field trials is discussed.
References
1. Bossanyi EA. Electrical aspects of variable speed operation of horizontal axis wind turbine generators. W/33/00221/REP,
ETSU, Harwell, 1994.
2. Bossanyi EA. The design of closed loop controllers for wind turbines. Wind Energy 2000; 3: 149–163.
3. Holley W, Rock S, Chaney K. Control of variable speed wind turbines below rated wind speed. Proceedings of 3rd
ASME/JSME Conference, 1999.
4. Bossanyi E. Bladed User Manual . Garrad Hassan: Bristol, 2002.
5. Caselitz P, Kleinkanf W, Krüger T, Petschenka J, Reichardt M, Störzel K. Reduction of fatigue loads on wind energy
converters by advanced control methods. Proceedings of European Wind Energy Conference, Dublin, 1997; 555–558.
6. Park RH. Two-reaction theory of synchronous machines. Transactions of the AIEE 1992; 48: 716–727.
7. Bossanyi E. Individual blade pitch control for load reduction. Wind Energy 2003; 6: 119–128.
8. Knudsen T, Andersen P, Töffner-Clausen S. Comparing PI and robust pitch controllers on a 400 kW Wind turbine by
full-scale tests. Proceedings of European Wind Energy Conferences, Dublin, 1997; 546–550.

Bossanyi 2003

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Bossanyi 2003

Uploaded by

Copyright:

Available Formats

WIND ENERGY

Wind Energ. 2003; 6:229–244 (DOI: 10.1002/we.95)

Review Wind Turbine Control for

Copyright  2003 John Wiley & Sons, Ltd.

Fixed-speed Pitch Controllers

Variable Speed—PI Torque Control

Variable Speed—Controlling Pitch and Torque Together

Effect of Turbulence on Peak Cp Tracking

ž rotor and generator rotation,

Controlling Tower Vibration

Drive Train Torsional Vibration

No damping With damping

Figure 2. Effect of a drive train damping filter

Measuring Tower Acceleration

No tower damping With tower damping

Individual Pitch Control

Blade 1 Blade 2 Blade 3 Collective

Blade root bending moment: out of plane

Shaft bending moment (My)

Yaw bearing yaw moment (Mz)

Figure 5. Load spectra

(a) Collective pitch controller

Figure 7. Reduction in fatigue loads

Evaluating Control Performance

ž drive train rotational and torsional dynamics;

The wind input should include:

ž wind shear, tower shadow and upflow;

You might also like